Complexity Beyond Incentives: The Critical Role of Reporting Language

2025-11-27

AI Summary

This study investigates the challenges of preference reporting in multi-attribute allocation (e.g., institutions, programs, tuition) and how interface design affects reporting accuracy and allocative welfare. Through a controlled lab experiment, we elicit preferences exhibiting lexicographic, separable, or complementary structures, and compare three reporting mechanisms—direct full ranking, attribute weighting, and sequential choice—across multiple interface designs. Results show that preference complexity induces substantial misreporting; interface simplification fails to improve accuracy; and sequential choice consistently outperforms traditional mechanisms across reporting accuracy, allocative efficiency, and fairness (reducing justified envy), while exhibiting greater robustness. This is the first systematic empirical demonstration of sequential choice’s superiority for multidimensional preference elicitation. The findings provide rigorous evidence and actionable design principles for real-world allocation systems, including college admissions and scholarship assignment.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Knowledge Representation and Reasoning: Preferences
• Multiagent Systems: Mechanism Design

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Many assignment systems require applicants to rank multi-attribute bundles (e.g., programs combining institution, major, and tuition). We study whether this reporting task is inherently difficult and how reporting interfaces affect accuracy and welfare. In laboratory experiments, we induce preferences over programs via utility over attributes, generating lexicographic, separable, or complementary preferences. We compare three reporting interfaces for the direct serial dictatorship mechanism: (i) a full ranking over programs; (ii) a lexicographic-nesting interface; and (iii) a weighted-attributes interface, the latter two eliciting rankings over attributes rather than programs. We also study the sequential serial dictatorship mechanism that is obviously strategy-proof and simplifies reporting by asking for a single choice at each step. Finally, we run a baseline that elicits a full ranking over programs but rewards pure accuracy rather than allocation outcomes. Four main findings emerge. First, substantial misreporting occurs even in the pure-accuracy baseline and increases with preference complexity. Second, serial dictatorship induces additional mistakes consistent with misperceived incentives. Third, simplified interfaces for the direct serial dictatorship fail to improve (and sometimes reduce) accuracy, even when they match the preference structure. Fourth, sequential choice achieves the highest accuracy while improving efficiency and reducing justified envy. These findings caution against restricted reporting languages and favor sequential choice when ranking burdens are salient.

Problem

• Studies difficulty of ranking multi-attribute bundles in assignment systems
• Compares reporting interfaces' effects on accuracy and welfare in experiments
• Evaluates sequential versus direct mechanisms for simplifying preference reporting

Innovation

• Sequential serial dictatorship mechanism for simplified reporting
• Lexicographic-nesting and weighted-attributes interfaces for attribute-based ranking
• Laboratory experiments inducing lexicographic, separable, or complementary preferences